haseebmalik18 commented on issue #62500: URL: https://github.com/apache/airflow/issues/62500#issuecomment-4032020644
I think a generation pipeline would be fragile long-term. Contributing docs are written for humans, while agent instructions need to be concise as Jason mentioned. A generator would need to decide what to keep vs. drop, which makes this a bit more complicated than a simple transformation. The more curated we want the instructions, the harder it is to automate reliably. What if we **maintain agent instructions by hand and validate them in CI**? **1. Command validation** If instructions follow a consistent format (e.g using code blocks for commands), CI can parse them and verify they still exist: - `breeze --help` checks subcommands - `uv run` paths must have a valid `pyproject.toml` - `pre-commit` hooks must exist in `.pre-commit-config.yaml` If something is renamed or removed, CI fails. **2. Identifier coverage** Add markers to workflow sections in the contributing docs. CI ensures every marker has a corresponding section in the agent instructions. If a new workflow is added without coverage, CI fails. **3. Codeowners review** If workflow logic changes without adding commands or workflows, the first two checks won’t catch it. A `CODEOWNERS` rule requiring approval from the agent-instructions owner ensures these subtle changes are reviewed. The first two checks are fully automated and catch most issues. The third covers subtle logic changes that are hard to detect automatically, so it requires a human review. @jason810496 What do you think? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
